考虑性能退化的航空发动机故障诊断量化评估

Quantitative Evaluation of Fault Diagnosability for Degraded Aero-Engine

关键词:  航空发动机  故障可诊断性  量化评价  性能退化  巴氏距离  最优状态观测器  距离相似度

Key words:  Aero-engine  Fault diagnosability  Quantitative evaluation  Performance degradation  Bhattacharyya coefficient  Optimal state observer  Distance similarity

摘要:

航空发动机全生命周期内的故障诊断一直是研究的热点，为保障航空发动机在性能退化条件下故障诊断算法的可靠性，对考虑性能退化的航空发动机进行故障可诊断性量化评估有着重要的意义。从状态量的可测量性以及最优观测器设计两个方面对退化状态下航空发动机的滑动窗口模型进行解耦处理。从故障可检测及可隔离两个方面对故障的可诊断性进行量化评估，以巴氏距离为量化标准，将量化评估问题转换为多元分布概率距离求解问题。同时，从故障空间的角度对故障可隔离性进行定义，剥离了参考故障模式的影响，将可隔离性转化为故障空间中故障模式的固有属性，并给出航空发动机故障可检测及可隔离的判据。仿真结果证明，本文所提出的方法可以在发动机性能退化条件下对控制系统传感器、执行机构故障以及发动机气路部件故障进行可诊断性量化评估。

Abstract:

Fault diagnosis during the entire life cycle of aero-engines has always been a hot research topic. In order to ensure the reliability of the aero-engine fault diagnosis algorithm under the condition of performance degradation， it is of great significance to quantitatively evaluate the fault diagnosability of aero-engines considering the performance degradation. Decouple the sliding window model of degraded areo-engine from two aspects： the measurability of the state quantity and the design of optimal state observer. The diagnosability is evaluated from two perspectives： detectability and isolability， then introducing Bhattacharyya distance as quantitative standard， and converting quantitative evaluation of fault diagnosability to calculation of multivariate distribution probability distance. Fault isolability is defined from the perspective of the fault space， the influence of the reference failure mode is stripped away， the isolability is transformed into the inherent attribute of the failure mode in the fault space， and the criterion for the detectability and isolability of aero-engine fault is provided. The simulation proves that the method proposed in this paper can perform diagnosability quantitative evaluation of the control system sensor， actuator failure and engine gas circuit component failure under the premise of aero-engine performance degradation.